Fast iterative decoding of linear dispersion codes for unknown MIMO channels

We propose an efficient iterative algorithm for demodulating/decoding linear dispersion (LD) codes in the case of an unknown MIMO channel. The algorithm is based on the POCS (projections onto convex sets) technique and consists of alternating projections of an estimate of the transmit matrix onto two linear subspaces. In the noise-free case, the iteration is guaranteed to converge to the transmitted data whenever the LD code theoretically allows perfect data reconstruction. In the presence of noise, simulation results demonstrate the algorithm´s good performance. We also propose a DFT-matrix based code design that yields a further substantial reduction of complexity, and a modified algorithm that exploits the presence of pilot symbols.